A new research project by the Journal of South Asian Linguistics has utilized deep-learning algorithms to trace 'loanwords' from Proto-Dravidian languages within 2nd-century CE Sanskrit botanical treatises. The study identifies a large cluster of terms related to specialized rice varieties and wetland irrigation that have no Indo-Aryan roots, suggesting they were absorbed from an advanced indigenous agricultural tradition.
The AI model reconstructed the phonetic evolution of these terms, showing that the Sanskrit authors were documenting an existing, highly sophisticated botanical classification system used by local farmers in the Deccan and Southern India. This research provides linguistic support for the archaeological theory that the 'Second Urbanization' of India was a deeply collaborative process involving multiple linguistic and cultural groups.